The Role Of Posterior Parietal Cortex In Episode Encoding And Representation In Macaque Monkeys

A critical component of the adaptive brain is to process dynamic environmental inputs in real-time.Our brain system has to extract pieces of meaningful information,integrate and encode them into memory systems in time for unexpected needs.Yet the mechanisms for what kinds of environmental features contribute to the event formation,and how the evolving thread of content and passage of time embedded in episodic events are computed in the brain remain elusive.To address this,we recorded single unit activities from the posterior parietal cortex(PPC)of three monkeys with a 32-ch micro-drive while they underwent a natural videoviewing experiment.In each session,the monkeys viewed three different 30-s videos,with each one repeated 30 times.We summarized main results based on analysis from single unit and ensembles as follows:To test whether PPC neurons process content-related activities in the movies and how neuronal dynamics change over repetitive exposures,we compared firing-rates across the three conditions for each neuron.About 33% PPC neurons respond selectively to categorical video-types,among which 20.27%,6.67% and 5.6% neurons being sensitive to primate,non-primate and scenery contents,respectively.Interestingly,23.2% and 26.67% of the neurons decreased and increased its firing rate over repetitions respectively,showing repetition suppression and repetition enhancement.To examine to what extent these content-sensitive neurons differentiate the three classes of videos,we applied leave-one-out cross-validation SVM classification with1-s time bin on the firing rates.Finally,55.2% PPC neurons predicted specific videos with higher accuracy than chance level.Especially,primate-sensitive neurons successively discriminated primate content over non-primate and scenery content,while other two failed to separate corresponding contents from others.Our results implied that PPC neurons may carry categorical information,especially for conspecifics activities.We extract low-level visual features to explore how these features influence neuronal encoding.Based on comparisons among videos for each feature,results revealed that neuronal decodability profit from contrast and optic flow.Importantly,motion-related neurons with higher predictive ability to primate videos,indicating that conspecifics activities installed in PPC neuronal encoding.Ethograms of video high social feature were labelled frame by frame for feature selection using LASSO model to investigate what specifics were encoded during movie watching.Results showed that the visible genital contribute to primate-sensitive neurons encode the movie content.To examine how voluntary attention tunes PPC neurons encode video content,we labelled fixation target.Results demonstrated that monkey fixed more for conspecifics,moreover,neuronal firing more to conspecifics' body,showing that PPC neurons represent conspecifics activities in macaque monkeys.Finally,to approach how these low-level and high-level features were temporally built up into events,accumulated spiking sequences were applied to train the SVM classifier for single neuron.Results demonstrated that prediction accuracies increased along with the neuronal temporal information accumulation.SVM classifier was also trained using ensemble activities to decode 1s epoch for each video.Results indicated that population activities segmented 1s epochs from each other,especially for conspecifics content.All in all,our study evidenced that PPC neurons temporally integrate low-level and high-level features into events at both individual and ensemble levels during free viewing.